Climate variability has a strong impact on maize yield. For example, the strong drought that occurred in 2016 led to lower yields across France, even for irrigated fields. Yield estimates have a significant strategic and economic importance. High spatial and temporal resolution remote sensing data are a valuable tool for providing yield estimates at a large scale.
In a recent study (Battude et al. 2016) based on optical image time series (combination of Formosat-2, Landsat-8, SPOT4-Take5 and Deimos-1, about two images per month), CESBIO researchers have developed a new method for the estimation of maize yield. A new formulation of SAFY agro-meteorological model taking into account of the observed seasonal variation of the specific leaf area (SLA) and the effective light use efficiency (ELUE) was proposed.
Results show that these modifications improve biomass estimates at local scale.
Comparison of measured and simulated Dry Aboveground Mass (DAM) with the original version of SAFY (left) and the new model version (right)
Yield estimates are compared to annual statistical values (Agreste) on two departments in the southwest of France : the Gers and the Haute-Garonne. Results show that the model reproduces well yields (R = 0.96; RRMSE = 4.6%), even if it sometimes overestimates the values for rainfed fields.
Comparison of simulated yield and Agreste values [t.ha-1] for the Gers and Haute-Garonne departments in 2013 (left) and 2014 (right), with the distinction between irrigated and rainfed fields. Standard errors associated to simulated values are reported.
GAI thus seems to be a good indicator for estimating the irrigated maize yield at regional scale. For rainfed fields, coupling SAFY with a water balance module simulating the soil water content may improve yield estimates. Sentinel-2 mission offers new perspectives and its data should improve the model estimates.
Reference : Battude M., Al Bitar A., Morin D., Cros J., Huc M., Marais Sicre C., Le Dantec V., Demarez V. (2016) Estimating maize biomass and yield over large area using high spatial and temporal resolution Sentinel-2 like remote sensing data. Remote Sensing of Environment 184, 668-681 DOI: 10.1016/j.rse.2016.07.030